rc4-mm1 and pwc-unofficial: kernel BUG and scheduling while atomic

rc4-mm1 and pwc-unofficial: kernel BUG and scheduling while atomic

[Norbert Preining]

Hi Luc, hi Andrew, hi list!

I tried to get the unofficial pwc modules from
http://www.saillard.org:8080/pwc/ to work with 2.6.9-rc4-mm1. After
fixing the following two things:
--- pwc-if.c.orig       2004-09-27 01:31:42.000000000 +0200
+++ pwc-if.c    2004-10-17 09:28:26.000000000 +0200
@@ -932,7 +932,7 @@
                if (urb != 0) {
                        if (pdev->iso_init) {
                                Trace(TRACE_MEMORY, "Unlinking URB %p\n", urb);
-                               usb_unlink_urb(urb);
+                               usb_kill_urb(urb);
                        }
                        Trace(TRACE_MEMORY, "Freeing URB\n");
                        usb_free_urb(urb);
@@ -1618,7 +1618,7 @@
        pos = (unsigned long)pdev->image_data;
        while (size > 0) {
                page = kvirt_to_pa(pos);
-               if (remap_page_range(vma, start, page, PAGE_SIZE,
                PAGE_SHARED))
+               if (remap_pfn_range(vma, start, page, PAGE_SIZE, PAGE_SHARED))
                        return -EAGAIN;
 
                start += PAGE_SIZE;

the module compiled and loaded without problem, but when starting
gnomemeeting I get the following kernel BUG and scheduling while atomic:

pwc Philips webcam module version 9.0.2-unofficial loaded.
pwc Supports Philips PCA645/646, PCVC675/680/690, PCVC720[40]/730/740/750 & PCVC830/840.
pwc Also supports the Askey VC010, various Logitech Quickcams, Samsung MPC-C10 and MPC-C30,
pwc the Creative WebCam 5 & Pro Ex, SOTEC Afina Eye and Visionite VCS-UC300 and VCS-UM100.
pwc Philips PCVC680K (Vesta Pro) USB webcam detected.
pwc Registered as /dev/video0.
usbcore: registered new driver Philips webcam
pwc type = 680
pwc type = 680
pwc set_video_mode(176x144 @ 10, palette 15).
pwc decode_size = 2.
pwc type = 680
pwc type = 680
pwc set_video_mode(176x144 @ 10, palette 15).
pwc decode_size = 2.
pwc type = 680
pwc type = 680
pwc set_video_mode(176x144 @ 10, palette 15).
pwc decode_size = 2.
pwc type = 680
pwc type = 680
pwc set_video_mode(176x144 @ 10, palette 15).
pwc decode_size = 2.
pwc type = 680
pwc type = 680
pwc set_video_mode(176x144 @ 10, palette 15).
pwc decode_size = 2.
------------[ cut here ]------------
kernel BUG at mm/rmap.c:468!
invalid operand: 0000 [#1]
PREEMPT 
Modules linked in: pwc snd_usb_audio snd_usb_lib nvidia snd_emu10k1 snd_rawmidi snd_seq_device snd_ac97_codec snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd_page_alloc snd_util_mem snd_hwdep snd soundcore via686a adm1021 i2c_sensor i2c_isa i2c_viapro i2c_core parport_pc lp parport
CPU:    0
EIP:    0060:[<c0145897>]    Tainted: P      VLI
EFLAGS: 00210286   (2.6.9-rc4-mm1) 
EIP is at page_remove_rmap+0x27/0x40
eax: ffffffff   ebx: 00001000   ecx: b4588000   edx: c12a0000
esi: df902624   edi: c12a0000   ebp: 00008000   esp: de18beb4
ds: 007b   es: 007b   ss: 0068
Process gnomemeeting (pid: 11671, threadinfo=de18a000 task=e40bc0c0)
Stack: c013f588 cc027080 00000000 e40bc0c0 c0116090 15000027 b4588000 c041d558 
       b4988000 cc027b48 b4590000 c041d558 c013f6cf 00008000 00000000 00000065 
       3b9a9e6f b4588000 cc027b48 b4590000 c041d558 c013f72d 00008000 00000000 
Call Trace:
 [<c013f588>] zap_pte_range+0x128/0x230
 [<c0116090>] default_wake_function+0x0/0x10
 [<c013f6cf>] zap_pmd_range+0x3f/0x60
 [<c013f72d>] unmap_page_range+0x3d/0x70
 [<c013f84e>] unmap_vmas+0xee/0x200
 [<c0143652>] unmap_region+0x72/0xe0
 [<c01438fd>] do_munmap+0xfd/0x150
 [<c0143990>] sys_munmap+0x40/0x70
 [<c0103fc9>] sysenter_past_esp+0x52/0x71
Code: 00 00 00 00 89 c2 8b 00 f6 c4 08 75 28 83 42 08 ff 0f 98 c0 84 c0 74 11 8b 42 08 40 78 0d 9c 58 fa ff 0d 30 1c 43 c0 50 9d 90 c3 <0f> 0b d4 01 21 f9 31 c0 eb e9 0f 0b d1 01 21 f9 31 c0 eb ce 90 
 <6>note: gnomemeeting[11671] exited with preempt_count 1
scheduling while atomic: gnomemeeting/0x00000001/11671
 [<c03076dd>] schedule+0x4dd/0x4f0
 [<c03080cd>] rwsem_down_read_failed+0x8d/0x190
 [<c012c78b>] .text.lock.futex+0x7/0xbc
 [<c012c64e>] do_futex+0x4e/0x80
 [<c012c760>] sys_futex+0xe0/0xf0
 [<c0119854>] release_console_sem+0xd4/0xe0
 [<c0117388>] mm_release+0x98/0xb0
 [<c011b2ca>] do_exit+0x7a/0x3f0
 [<c01051c0>] do_divide_error+0x0/0x120
 [<c0105520>] do_invalid_op+0x0/0xf0
 [<c0128c61>] search_exception_tables+0x21/0x30
 [<c0105520>] do_invalid_op+0x0/0xf0
 [<c010560b>] do_invalid_op+0xeb/0xf0
 [<c0145897>] page_remove_rmap+0x27/0x40
 [<f0e361a8>] pwc_decompress+0x228/0x2b0 [pwc]
 [<c0307499>] schedule+0x299/0x4f0
 [<c0104a49>] error_code+0x2d/0x38
 [<c0145897>] page_remove_rmap+0x27/0x40
 [<c013f588>] zap_pte_range+0x128/0x230
 [<c0116090>] default_wake_function+0x0/0x10
 [<c013f6cf>] zap_pmd_range+0x3f/0x60
 [<c013f72d>] unmap_page_range+0x3d/0x70
 [<c013f84e>] unmap_vmas+0xee/0x200
 [<c0143652>] unmap_region+0x72/0xe0
 [<c01438fd>] do_munmap+0xfd/0x150
 [<c0143990>] sys_munmap+0x40/0x70
 [<c0103fc9>] sysenter_past_esp+0x52/0x71

Thanks a lot for any information to fix this!

Best wishes

Norbert

-------------------------------------------------------------------------------
Norbert Preining <preining AT logic DOT at>         Technische Universität Wien
gpg DSA: 0x09C5B094      fp: 14DF 2E6C 0307 BE6D AD76  A9C0 D2BF 4AA3 09C5 B094
-------------------------------------------------------------------------------
SLUBBERY (n.)
The gooey drips of wax that dribble down the sides of a candle so
beloved by Italian restaurants with Chianti bottles instead of
wallpaper.
			--- Douglas Adams, The Meaning of Liff

Re: rc4-mm1 and pwc-unofficial: kernel BUG and scheduling while atomic

[William Lee Irwin III]

On Sun, Oct 17, 2004 at 09:36:14AM +0200, Norbert Preining wrote:
> @@ -1618,7 +1618,7 @@
>         pos = (unsigned long)pdev->image_data;
>         while (size > 0) {
>                 page = kvirt_to_pa(pos);
> -               if (remap_page_range(vma, start, page, PAGE_SIZE,
>                 PAGE_SHARED))
> +               if (remap_pfn_range(vma, start, page, PAGE_SIZE, PAGE_SHARED))
>                         return -EAGAIN;
>  
>                 start += PAGE_SIZE;
> the module compiled and loaded without problem, but when starting
> gnomemeeting I get the following kernel BUG and scheduling while atomic:

You need to right shift the argument by PAGE_SHIFT.


-- wli

Re: rc4-mm1 and pwc-unofficial: kernel BUG and scheduling while atomic

[William Lee Irwin III]

On Sun, Oct 17, 2004 at 09:36:14AM +0200, Norbert Preining wrote:
>> the module compiled and loaded without problem, but when starting
>> gnomemeeting I get the following kernel BUG and scheduling while atomic:

On Sun, Oct 17, 2004 at 02:30:18AM -0700, William Lee Irwin III wrote:
> You need to right shift the argument by PAGE_SHIFT.

We have handled this further in private email and Norbert now has
a working port of his driver.


-- wli

Re: rc4-mm1 and pwc-unofficial: kernel BUG and scheduling while atomic [u]

[Martin Schlemmer [c]]

[-- Attachment #1: Type: text/plain, Size: 1184 bytes --]

On Sun, 2004-10-17 at 02:30 -0700, William Lee Irwin III wrote:
> On Sun, Oct 17, 2004 at 09:36:14AM +0200, Norbert Preining wrote:
> > @@ -1618,7 +1618,7 @@
> >         pos = (unsigned long)pdev->image_data;
> >         while (size > 0) {
> >                 page = kvirt_to_pa(pos);
> > -               if (remap_page_range(vma, start, page, PAGE_SIZE,
> >                 PAGE_SHARED))
> > +               if (remap_pfn_range(vma, start, page, PAGE_SIZE, PAGE_SHARED))
> >                         return -EAGAIN;
> >  
> >                 start += PAGE_SIZE;
> > the module compiled and loaded without problem, but when starting
> > gnomemeeting I get the following kernel BUG and scheduling while atomic:
> 
> You need to right shift the argument by PAGE_SHIFT.
> 

I am trying to get vesafb-tng to work with rc4-mm1, but are not sure
when to shift the argument by PAGE_SHIFT, and when not to.  The patches
from you in rc4-mm1 sometimes shifts the second arg, other times the
third, and other times not at all.  Is there a easy way for a mostly
clueless person to figure out when to shift what argument and when not?


Thanks,

-- 
Martin Schlemmer


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: rc4-mm1 and pwc-unofficial: kernel BUG and scheduling while atomic [u]

[William Lee Irwin III]

On Sun, 2004-10-17 at 02:30 -0700, William Lee Irwin III wrote:
>> You need to right shift the argument by PAGE_SHIFT.

On Sun, Oct 17, 2004 at 08:35:31PM +0200, Martin Schlemmer [c] wrote:
> I am trying to get vesafb-tng to work with rc4-mm1, but are not sure
> when to shift the argument by PAGE_SHIFT, and when not to.  The patches
> from you in rc4-mm1 sometimes shifts the second arg, other times the
> third, and other times not at all.  Is there a easy way for a mostly
> clueless person to figure out when to shift what argument and when not?

Please point out where these inconsistencies occur and I will repair
them.

Only the third argument changed, from a physical address to a pfn.


-- wli

Re: rc4-mm1 and pwc-unofficial: kernel BUG and scheduling while atomic [u]

[Martin Schlemmer [c]]

[-- Attachment #1: Type: text/plain, Size: 1609 bytes --]

On Sun, 2004-10-17 at 12:00 -0700, William Lee Irwin III wrote:
> On Sun, 2004-10-17 at 02:30 -0700, William Lee Irwin III wrote:
> >> You need to right shift the argument by PAGE_SHIFT.
> 
> On Sun, Oct 17, 2004 at 08:35:31PM +0200, Martin Schlemmer [c] wrote:
> > I am trying to get vesafb-tng to work with rc4-mm1, but are not sure
> > when to shift the argument by PAGE_SHIFT, and when not to.  The patches
> > from you in rc4-mm1 sometimes shifts the second arg, other times the
> > third, and other times not at all.  Is there a easy way for a mostly
> > clueless person to figure out when to shift what argument and when not?
> 
> Please point out where these inconsistencies occur and I will repair
> them.
> 
> Only the third argument changed, from a physical address to a pfn.
> 

Its the vesafb-tng patch
(http://dev.gentoo.org/~spock/projects/vesafb-tng/)
Relevant part:

----
        vma.vm_mm = current->active_mm;
        vma.vm_page_prot.pgprot = PROT_READ | PROT_EXEC | PROT_WRITE;

        ret = remap_page_range(&vma, 0x000000, __pa(mem), REAL_MEM_SIZE, vma.vm_page_prot);
        ret += remap_page_range(&vma, 0x0a0000, 0x0a0000, 0x100000 - 0x0a0000, vma.vm_page_prot);
----

I did it as:

----
        vma.vm_mm = current->active_mm;
        vma.vm_page_prot.pgprot = PROT_READ | PROT_EXEC | PROT_WRITE;

        ret = remap_pfn_range(&vma, 0x000000, __pa(mem) >> PAGE_SHIFT, REAL_MEM_SIZE, vma.vm_page_prot);
        ret += remap_pfn_range(&vma, 0x0a0000, 0x0a0000, 0x100000 - 0x0a0000, vma.vm_page_prot);
----


Thanks,
-- 
Martin Schlemmer


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: rc4-mm1 and pwc-unofficial: kernel BUG and scheduling while atomic [u]

[William Lee Irwin III]

On Sun, 2004-10-17 at 12:00 -0700, William Lee Irwin III wrote:
>> Only the third argument changed, from a physical address to a pfn.

On Sun, Oct 17, 2004 at 09:57:45PM +0200, Martin Schlemmer [c] wrote:
> Its the vesafb-tng patch
> (http://dev.gentoo.org/~spock/projects/vesafb-tng/)
> I did it as:
> 
> ----
>         vma.vm_mm = current->active_mm;
>         vma.vm_page_prot.pgprot = PROT_READ | PROT_EXEC | PROT_WRITE;
> 
>         ret = remap_pfn_range(&vma, 0x000000, __pa(mem) >> PAGE_SHIFT, REAL_MEM_SIZE, vma.vm_page_prot);
>         ret += remap_pfn_range(&vma, 0x0a0000, 0x0a0000, 0x100000 - 0x0a0000, vma.vm_page_prot);

You probably have to shift the physical address in the second call also.


-- wli